Screening and mechanism analysis of negative thermal expansion in Bi_0.85Ca_0.15Fe_0.85M_0.15O₃ (M: tetravalent cations) via machine learning force fields

Abstract

Negative thermal expansion (NTE) is a key property for achieving dimensional stability in advanced technologies, yet the discovery of new NTE materials is often hindered by the vast compositional space and the complexity of phase transition mechanisms. In this study, we applied large-scale molecular dynamics simulations using the Crystal Hamiltonian Graph Network (CHGNet), a pretrained universal machine learning force field (MLFF), to screen for NTE candidates in the Bi_0.85Ca_0.15Fe_0.85M_0.15O₃ system (M = Ti, Zr, Sn, Hf, Ir, Pb, Ru). Our systematic simulations reveal that the co-substitution of Ca and tetravalent cations effectively modulates the phase stability of BiFeO₃, with the transition temperature varying depending on the substituting element. Notably, the Ir-substituted compound exhibits a sharp volume-collapsing first-order phase transition at lower temperature (∼380 K) with a substitution level of 15 % in good agreement with the experimental result. These findings demonstrate that universal MLFFs are powerful tools for accelerating the exploration of functional materials and clarifying atomic-level mechanisms of complex phase transitions.